-
Notifications
You must be signed in to change notification settings - Fork 0
FlexDR Documentation
Welcome to the FlexDR backend wiki!
This section aims to provide details of the FlexDR backend implementation focusing on the business perspective of the service.
More information regarding the FlexDR service could be found in the following links:
- I-NERGY FlexDR frontend repository documentation
- I-NERGY FlexDR orchestration engine
- I-NERGY project's website
Meters refer to the smart meters involved in the energy community. These smart meters along with their details could be easily registered via the FlexDR UI.
ML model refers to the clustering algorithm used to detect the load profiles of the smart meters involved in the energy community within I-NERGY UC7. ML models contain all the information regarding the algorithm and its configuration used for clustering (e.g. kmeans, 14 clusters, euclidean distance) as well as the identified clusters in form of load curves. The ML model is tracked within I-NERGY MLflow instance, and is deployed as a service. ML model is registered via the API within FlexDR backend and MongoDB instance.
The concept behind a load profile is to group together multiple clusters identified by the ML model. If similarities exist between these profiles, Energy Experts, based on their expertise, can simplify and merge the clusters for better management and analysis. Load profile profile is created via the FlexDR UI and is linked with an existing ML model. It contains user defined attributes, such as description and a relevant recommendation, and a single or multiple clusters which is/are a subset of the clusters of the ML model.
Assignments refer to the forecasts generated by FlexDR service. This entity includes the following information:
- The smart meter for which the FlexDR service generate the forecast
- The ML model instance that is used to generate the forecast
- The forecast data, including both the forecast load curve and the predicted matched cluster (load curve) according to the ML model. Assignments creation is an automated process within I-NERGY FlexDR process and is carried out by FlexDR's workflow orchestration engine. This automated job is responsible for performing the following tasks:
- Load smart meters data from I-NERGY uc7 database
- Preprocess data
- Request day ahead load forecasts from the inference endpoint of the DeepTSF service
- Request matched clusters for the forecast loads using the inference endpoint of the FlexDR's clustering service
- Generate and register the appropriate assignments according to previous two steps within FlexDR mongo db.